Inception Nesterov Momentum Adam L2 Regularized Learning Rate CNN for Sugarcane Disease Classification

The primary source of sugar extractor is sugarcane, which is regarded as a revenue source crop globally. Sugarcane is a cash crop that greatly benefits the economy. Therefore, it is crucial for farmers to identify and cure diseases in the sugarcane production at the appropriate time. To address this issue, deep learning techniques are employed. The disease in the sugarcane crop could affect the overall yield of the crop. With this view, this paper recommends Inception Nadam L2 Regularized Gradient Descent (NLRGD) CNN model to classify the sugar cane disease like red rot, smut, mosaic, yellow leaf, red-orange rust using the Sugarcane Disease Dataset from KAGGLE. The proposed NLRGD model uses Sugarcane Disease Dataset from KAGGLE with 3000 sugarcane plant leaflets having 1000 Bacterial Blight, 1000 Red Dot and 1000 healthy sugarcane plant images. The CXAGD model was designed using 3 X 3 convolutional layers and maximum pooling layer followed by 1 X 1 Dropout and Dense layer for each 3 X 3 convolution and trained with learning rate of L2 regularized Gradient Descent and NADAM optimizer. The NLRGD built with Inception network with L2 regularized Gradient Descent and NADAM optimizer with 0.5 dropout, weight decay of Gradient Descent L2 regularization. The dataset for sugarcane disease was partitioned into training and testing data, and the training data were fitted to both the NLRGD model and other CNN models in order to compare the effectiveness of the proposed NLRGD model. Python was coded on a Gtx Geforce Tesla V100 Nvidia graphics card server with a batch size of 64 and 40 training epochs. Results of execution indicate that the suggested NLRGD model showcase the accuracy of 96.75%, Precision of 96.62%, Recall of 96.25% and FScore of 96.76% when compared with existing CNN models.